CN-121984974-A - Intelligent computing center distributed cooperative processing method and system based on edge calculation

CN121984974ACN 121984974 ACN121984974 ACN 121984974ACN-121984974-A

Abstract

The invention belongs to the technical field of intelligent computation center scheduling, and particularly relates to an intelligent computation center distributed cooperative processing method and system based on edge computing. The method comprises the steps of evaluating the real-time running state of each distributed computing node to identify available nodes, classifying distributed computing tasks based on task content, which needs to be interacted among the nodes, in the distributed computing tasks, distributing the distributed computing tasks to the available nodes for execution according to task classification results, determining execution correctness according to feedback results, and redistributing the distributed computing tasks with execution errors to be executed when determining the distributed computing tasks with the execution errors. The invention reduces the failure rate of the tasks, improves the resource allocation efficiency and the utilization rate and enhances the fault-tolerant self-healing capacity of the system by pre-evaluating the nodes, dynamically classifying the tasks and performing closed-loop error correction, thereby guaranteeing the stability, the high efficiency and the service continuity of the distributed cooperative processing.

Inventors

YANG ZHEN

Assignees

北京安联通科技有限公司
杨桢

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (9)

1. An intelligent computing center distributed cooperative processing method based on edge calculation is characterized by comprising the following steps: The method comprises the steps of evaluating real-time running states of distributed computing nodes subordinate to an intelligent computing center to identify available nodes, classifying distributed computing tasks based on task content needing to be interacted among nodes in the distributed computing tasks to be processed, outputting task classification results, distributing the distributed computing tasks to the available nodes to be executed according to the task classification results, redistributing the distributed computing tasks with execution errors to be executed when the distributed computing tasks with execution errors are determined according to feedback results generated by the distributed computing nodes executing the distributed computing tasks, and evaluating the real-time running states of the distributed computing nodes subordinate to the intelligent computing center to identify the available nodes.
2. The intelligent computing center distributed collaborative processing method based on edge computing according to claim 1, wherein the step of identifying available nodes based on real-time operation state information, response delay data, temperature conditions and historical operation data comprises judging failure probability of each distributed computing node according to the historical operation data, judging that the distributed computing node is an unavailable node when the failure probability exceeds a preset failure probability threshold, and when the failure probability does not exceed the preset failure probability threshold, performing temperature trend analysis by using historical temperature data in the historical operation data and combining current temperature data acquired from the temperature conditions, and judging whether the distributed computing node is the available node according to the result of the temperature trend analysis, wherein if the current temperature data falls into a preset temperature risk interval, judging that the distributed computing node is the unavailable node.
3. The intelligent computing center distributed collaborative processing method based on edge computing according to claim 1, wherein the step of classifying distributed computing tasks based on task content that requires inter-node interaction in the distributed computing tasks to be processed includes generating an interaction list containing inter-node interaction relationships to be checked according to task content that requires inter-node interaction, analyzing the interaction list to extract classification parameters, and classifying the distributed computing tasks based on the classification parameters.
4. The method for distributed collaborative processing in an intelligent computing center based on edge computing according to claim 3, wherein the step of analyzing the interaction list and extracting the classification parameters includes obtaining sub-content and data flow of the distributed computing task related to the interaction relationship based on the interaction relationship in the interaction list, converting the data amount of the sub-content into numerical parameters, and operating the numerical parameters to generate normalized ratio values as the classification parameters.
5. The method of claim 3, wherein classifying the distributed computing tasks based on classification parameters comprises determining classification thresholds according to load ranges and adaptive tolerances of the distributed computing nodes among the available nodes, comparing the classification parameters with the classification thresholds, and classifying the distributed computing tasks as synchronous type tasks or asynchronous type tasks based on the comparison results to generate task classification results.
6. The method for distributed collaborative processing in an intelligent computing center based on edge computing according to claim 1, wherein determining that there is an erroneous distributed computing task based on feedback results generated by distributed computing nodes executing the distributed computing task includes integrating the feedback results to generate associated parameters, performing a difference judgment on the associated parameters to determine the correctness of the execution of the distributed computing task, and determining whether there is an erroneous distributed computing task based on the result of the determination of the correctness of the execution.
7. The method according to claim 1, wherein the step of assigning the distributed computing tasks to the available nodes according to the task classification result comprises preferentially assigning the synchronous class tasks in the task classification result to the plurality of distributed computing nodes for synchronous execution, and the step of reassigning the erroneous distributed computing tasks for execution comprises asynchronously assigning the erroneous distributed computing tasks to the distributed computing nodes selected from the available nodes for execution.
8. An edge-computing-based distributed collaborative processing system for a computing center, comprising: The node state evaluation module is used for evaluating the real-time running state of each distributed computing node subordinate to the intelligent computing center so as to identify available nodes; the task classification module is used for classifying the distributed computing tasks based on task content which needs to be interacted among nodes in the distributed computing tasks to be processed so as to output task classification results; the execution correctness determining module is used for determining whether the execution error distributed computing task exists according to feedback results generated by all the distributed computing nodes executing the distributed computing task; and the task allocation module is used for allocating the distributed computing tasks to the available nodes for execution according to the task classification result, and when the execution correctness determination module determines that the distributed computing tasks with the execution errors exist, the task allocation module is also used for reallocating the distributed computing tasks with the execution errors for execution.
9. An intelligent computing center distributed cooperative processing terminal based on edge calculation, which is characterized by comprising: at least one processor, and a memory communicatively coupled to the at least one processor; Wherein the memory stores instructions executable by the at least one processor, which when executed, cause the at least one processor to perform a distributed co-processing method for an edge-based computing center as claimed in any one of claims 1 to 7.

Description

Intelligent computing center distributed cooperative processing method and system based on edge calculation Technical Field The invention belongs to the technical field of intelligent computation center scheduling, and particularly relates to an intelligent computation center distributed cooperative processing method and system based on edge computing. Background The intelligent computing center is used as a new generation information infrastructure for carrying mass data processing and complex model training, huge computing tasks are reasonably distributed to hundreds of distributed computing nodes in the intelligent computing center, so that the intensive utilization of computing resources and the maximization of the overall computing efficiency are realized, and therefore, the operation efficiency and the service capacity of the intelligent computing center are determined by the advantages and disadvantages of a task scheduling strategy. In the prior art, when dealing with the current large-scale and high-dynamic computing scene, the problems of hysteresis of scheduling decisions and inaccuracy of state information are prominent, a centralized main control server needs to periodically collect state information from all computing nodes, communication overhead and time delay exist, when the node scale is huge, the node state acquired by the main control server is often outdated, the scheduling decisions made based on the node state is not optimally matched with real-time resource conditions, the task execution efficiency and the system throughput are affected, an evaluation model of the node execution capacity by the existing scheduling method is too simple, the resource mismatch and the utilization rate are low, the traditional scheduling logic is mostly dependent on static or macroscopic indexes such as CPU occupancy rate, memory allowance and the like, the actual bearing capacity of the node for a specific type of computing task cannot be accurately depicted, the task is easily distributed to the non-optimal node, the robustness and the expandability of the system are limited, the master-slave architecture is highly dependent on the stability of the main control server, and the communication and the computing pressure of the central server are exponentially increased with the continuous increase of the number of computing nodes, and the system is difficult to smoothly and transversely expand. In view of this, the application provides a distributed collaborative processing method and system for an intelligent computing center based on edge computing, which overcomes the problems of delay, inefficiency and vulnerability caused by a centralized architecture, and realizes the accurate evaluation of the real-time sensing and execution capacity of the computing node state, thereby improving the resource allocation efficiency, the system robustness and the overall service level of the intelligent computing center. Disclosure of Invention The invention aims to provide an intelligent computing center distributed cooperative processing method and system based on edge calculation, which are used for solving the problem that the actual execution capacity of a computing node is difficult to accurately judge in the prior art, so that certain hysteresis exists in the scheduling work of the computing node. In order to achieve the aim of the invention, the technical scheme adopted by the invention is that the intelligent computation center distributed cooperative processing method based on edge computation comprises the following steps: Evaluating the real-time running state of each distributed computing node subordinate to the intelligent computing center to identify available nodes; classifying distributed computing tasks based on task content which needs to be interacted among nodes in the distributed computing tasks to be processed, so as to output task classification results; distributing the distributed computing tasks to available nodes for execution according to the task classification result; When determining that the distributed computing task with the execution error exists according to feedback results generated by each distributed computing node executing the distributed computing task, reassigning the distributed computing task with the execution error to execute; the step of evaluating the real-time operational status of each distributed computing node subordinate to the intelligent computing center to identify available nodes includes: Acquiring real-time running state information, response delay data, temperature conditions and historical running data of each distributed computing node in a preset time period; based on the real-time operational status information, the response delay data, the temperature conditions, and the historical operational data, available nodes are identified. Preferably, the step of identifying available nodes based on real-time operational status information, response delay dat